Jason O'Kane: Past Research Projects

These are research projects that have been (mostly) dormant for a few years.

Descriptions of more recent projects appear here.

Comparing Robot Systems

[illustration] [illustration]

Real robots must effectively collect, interpret, and act upon sensor data. How sophisticated must a robot's sensors be to complete a given task? What are the neccessary conditions? We seek the simplest robots that can complete a task, giving a precise meaning to the idea of simpleness. My goal is to develop a clean, formal technique for comparing robot systems and for studying their ability to complete tasks of varying difficulty. This work draws inspiration from the theory of computation, which plays a similar role in the core of computer science.

Localization with Minimal Sensing

Localization, the task of determining a robot's position within its environment, is one of the most important problems for mobile robots. How hard is this problem, in terms of the sensing and motion ability needed to complete it? This work is an investigation of the information requirements of the localization task. We found upper and lower bounds on the sensing and motion capabilities needed for localization and implemented our algorithms on Roomba vacuum cleaner robots.

Pareto Optimal Coordination

[illustration]
[illustration] In many situations, teams of robots must interact in a shared workspace. If the robots have distinct goals and objective functions, then the design of collision-free coordination plans for these teams is a multi-objective optimization problem. This work uses the idea of Pareto optimality to generate a set of non-dominated solutions to multiple-robot coordination problems.